Journal of the Korean Physical Society, Vol. 56, No. 2, February 2010, pp. 672 676 Structural Optimization of Silicon Carbide PIN Avalanche Photodiodes for UV Detection Ho-Young Cha School of Electronic and Electrical Engineering, Hongik University, Seoul 121-791 (Received 7 December 2009, in final form 23 December 2009) The locally-enhanced electric field occurring at the etched junction surface of mesa-etched PIN avalanche photodiodes (APDs) causes early edge breakdown, resulting in a relatively lower multiplication and gain. The electric field distribution as a function of the intrinsic layer thickness and the bevel angle was investigated, and its influence on the characteristics of 4H-SiC PIN APDs, such as the breakdown voltage, the quantum efficiency, and the gain, was studied. In addition, a new structure with a field plate is proposed to diminish the edge breakdown phenomenon in mesa-etched PIN APDs. PACS numbers: 85.60.Dw, 85.60.Bt, 81.05.Hd Keywords: Avalanche photodiode, Edge breakdown, Field plate, Silicon carbide DOI: 10.3938/jkps.56.672 I. INTRODUCTION Solid-state UV detectors have received great attention over photomultiplier tubes because of the benefit of their being cheap and portable devices. Currently, UV-enhanced Si photodetectors are commercially available for UV detection [1,2]. However, their low signalto-noise ratio in the deep-uv region, device aging, and sensitivity to visible photons are inherent limitations in practical applications [3]. In this point of view, silicon carbide (SiC) is a suitable candidate for the detection of UV radiation because it has high optical absorption in the UV range and exhibits a very low level of dark current due to its wide energy bandgap [3 5]. In addition, SiC devices can be utilized for high-temperature applications whereas other photodetectors require an extra cooling unit. Potential applications of UV photodetectors include jet engine/missile plume detectors, biological agent detection, etc [6]. PIN avalanche photodiodes (APDs) are a popular type of solid-state detector due to the enhanced signal-tonoise ratio. In order to achieve PIN APDs with high multiplication and gain, a uniform distribution of high electric field in the intrinsic region at the operating reverse bias is needed. When the active region in a photodiode is defined by a mesa etching process, a locally-enhanced electric field occurs at the etched junction surface, causing an early edge breakdown that depends on the bevel profile. It becomes a severe problem if the lateral extension of the highly-doped region is smaller than that of E-mail: hcha@hongik.ac.kr; Fax: +82-2-320-1119 a lowly-doped region at the junction [7,8]. One way to suppress the locally-enhanced electric field at the junction surface is to implement a positive bevel angle in the etched surface. However, that is not a solution for typical PIN structures because the upper junction between the doped and the intrinsic layers has a negative bevel profile whereas the lower junction has a positive one. As a result, the locally-enhanced electric field at the upper junction cannot be suppressed completely. If the electric field at the junction surface is higher than that in the active intrinsic region, an avalanche process will be initiated at the surface region; thus, the multiplication process in the active region will be limited by the lower electric field strength, resulting in a poor signal-to-noise ratio. In this study, we not only investigated the electric field distribution in 4H-SiC PIN APDs as a function of the intrinsic layer thickness and the bevel angle, but also proposed a new structure to suppress the high electric field at the surface. II. STRUCTURAL ANALYSIS AND OPTIMIZATION The structure of SiC PIN APD consisted of a P + anode layer, an N intrinsic layer, an N cathode layer, and an N + contact layer, as shown in Fig. 1. The doping concentration and the thickness of each layer are summarized in Table 1, where the intrinsic layer thickness t is the dimensional variable of interest. The doping concentration of each layer was determined, considering the feasibility of material growth. Using N-type materials for -672-
Structural Optimization of Silicon Carbide PIN Avalanche Photodiodes for UV Detection Ho-Young Cha -673- Fig. 1. Schematic cross-sectional view of a 4H-SiC PIN APD (t: intrinsic layer thickness and θ: bevel angle). Table 1. Doping concentration and thickness of each layer in the 4H-SiC PIN APD. Doping concentration (/cm 3 ) Thickness (nm) N + contact layer 2 10 19 100 N cathode layer 2 10 18 200 N intrinsic layer 1 10 16 600 < t < 3000 P + anode layer 3 10 18 2000 the upper layers is beneficial as the diffusion length of minority holes in N-type 4H-SiC has been reported to be longer than that of minority electrons in P-type 4H-SiC [9,10], which means that more carriers can diffuse into the intrinsic region, giving rise to higher photocurrents, particularly at high photon energies where the photon penetration depth is very shallow [11]. The influence of the intrinsic layer thickness t and the bevel angle θ on the device characteristics was investigated using a comprehensive two-dimensional device simulator provided by SILVACO. Detailed information on the physical models and the material parameters used for simulation can be found elsewhere [6]. Increasing the intrinsic layer thickness in a PIN structure generally leads to a lower dark current and a higher breakdown voltage. In addition, because the intrinsic region is where the high electric field is applied under high reverse bias conditions, more photons can participate in the multiplication process with a thicker intrinsic layer, which, in turn, results in a high gain and signal-to-noise ratio. However, such enhancement will be limited, depending on the bevel angle, which will be discussed below. The intrinsic layer thickness and the bevel angle were varied from 600 nm to 3000 nm and from 10 to 80, respectively. The behaviors of the breakdown voltage as functions of the bevel angle for various intrinsic layer thicknesses are shown in Fig. 2(a). The breakdown voltage of the SiC PIN APD with a 600-nm intrinsic layer is independent of the bevel angle whereas that with thicker intrinsic layers exhibits a noticeable dependency on the bevel angle. The breakdown voltage decreases as the bevel angle approaches 35 45 and is more significant Fig. 2. (a) Breakdown voltage vs the bevel angle for various intrinsic layer thicknesses. (b) Maximum electric field in the active intrinsic region vs the intrinsic layer thickness for various bevel angles. The reverse biases were set to the corresponding breakdown voltages shown in (a). with increasing intrinsic layer thickness, for example, a 20% reduction for a 3000-nm-thick intrinsic layer. The maximum electric field in the active intrinsic region was examined with the reverse bias set to the breakdown voltages obtained in Fig. 2(a). It is obvious that the maximum electric field decreases as the intrinsic layer thickness increases. Therefore, we suggest that the impact ionization is initiated not in the active region, but elsewhere. According to the electric field distribution simulated with the given structures, the highest electric field was, indeed, found at the upper junction on the etched surface, which is responsible for the early edge breakdown phenomenon. Consequently, incident photons in a SiC PIN APD with a thick intrinsic layer do not experience strong avalanche multiplication prior to the breakdown, which, in turn, will result in a poor gain. The quantum efficiency was also investigated as a function of the intrinsic layer thickness. Because the quan-
-674- Journal of the Korean Physical Society, Vol. 56, No. 2, February 2010 Fig. 4. Schematic cross-sectional view of a 4H-SiC PIN APD with a field plate. Fig. 3. Quantum efficiency vs wavelength for various intrinsic layer thicknesses. The reverse bias was set to 50 V, which is far below the breakdown region. tum efficiency was calculated at a low reverse bias (i.e. 50 V) far below the breakdown voltage, it is not affected by the early edge breakdown and is, thus, independent of the bevel angle. The quantum efficiency versus wavelength for various intrinsic layer thicknesses is shown in Fig. 3. A maximum quantum efficiency of 80% was obtained at wavelengths of 260 280 nm, which is much superior to what was achieved using a SiC separate absorption and multiplication APD (SAM-APD) [6]. A detailed discussion of SiC SAM-APDs can be found elsewhere [11]. In Fig. 3, the quantum efficiency is relatively independent of the intrinsic layer thickness in the short wavelength region, which is associated with a significant increase in the surface reflection and a decrease in the penetration depth due to the high photon energy. On the other hand, the quantum efficiency at wavelengths >250 nm increases with increasing intrinsic layer thickness, which is due to more photons being available in the intrinsic region. There are potential applications utilizing wavelengths in the range of 270 300 nm, such as non-line-of-sight covert communication and specific bioagent detection [12,13]. In order to achieve a high quantum efficiency in this wavelength region, it is necessary to employ a thick intrinsic layer. However, such a thick PIN structure cannot be used in an avalanche mode due to the significant reduction in multiplication caused by the limited electric field strength in the active intrinsic region, as discussed above. In order to solve this inherent problem, the maximum electric field in the active intrinsic region must be increased without an early edge breakdown. A field plate structure was incorporated into the mesa- Fig. 5. Dark current, photocurrent, and gain vs reverse bias voltage: (a) mesa-etched 4H-SiC PIN APD without a field plate, and (b) mesa-etched 4H-SiC APD with a field plate. Both have an intrinsic layer thickness of 2700 nm and a bevel angle of 10. etched SiC PIN APD to suppress the locally-enhanced electric field at the junction on the etched surface. A schematic cross-sectional view of the field plate PIN APD is shown in Fig. 4. The field plate was placed on the upper edge of the mesa, was separated by a 1-µm-thick oxide, and was extended laterally over the upper junction region. The target wavelength of interest was 280 nm,
Structural Optimization of Silicon Carbide PIN Avalanche Photodiodes for UV Detection Ho-Young Cha -675- Fig. 6. Electric field distribution of a 4H-SiC PIN APD without a field plate: (a) two-dimensional view and (b) threedimensional view. The reverse bias was set to 515 V (breakdown voltage). Fig. 7. Electric field distribution of a 4H-SiC PIN APD with a field plate: (a) two-dimensional view and (b) threedimensional view. The reverse bias was set to 585 V (breakdown voltage). and a 2700-nm-thick intrinsic layer was chosen to achieve a high photoresponsivity. Two structures with the same bevel angel (10 ) were compared: one without a field plate (conventional structure) and the other one with a field plate (field plate structure). The current-voltage characteristics of the two structures with and without UV illumination are compared in Fig. 5 where the device s active area is 100 µm 2. The breakdown voltage of the conventional structure is much lower than that of the field plate structure due to the early edge breakdown. A light source with a wavelength of 280 nm was used to illuminate the top of the active region. The incident photons in the conventional structure experience little multiplication even at a reverse bias near the breakdown voltage because of the relatively low electric field strength in the active intrinsic region. On the other hand, a much higher electric field is applied to the active region for the field plate structure; thus, the incident photons participate in a strong avalanche process. As a result, high gain can be achieved in the field plate structure, which is clearly seen in the electric field distribution simulated under a reverse bias near the breakdown voltage. Twodimensional and three-dimensional views of the electric field distribution for both structures are shown in Figs. 6 and 7. The electric field in the active intrinsic region of the field plate structure ranges from 1.9 MV/cm to 2.4 MV/cm whereas that of the conventional structure ranges from 1.6 MV/cm to 2.1 MV/cm. The enhanced electric field strength in the field plate structure resulted in superior detection characteristics. III. CONCLUSION The locally-enhanced electric field on the etched mesa surface causes an early edge breakdown in 4H-SiC PIN structures with a thick intrinsic layer and, thus, limits the maximum electric field in the active intrinsic region. This limited electric field strength results in poor multiplication and gain. Such degradation becomes an important issue as the intrinsic layer thickness increases. A field plate was incorporated into a conventional 4H-SiC PIN APD to suppress the high electric field at the etched surface. As a result, the maximum electric field in the active intrinsic region was noticeably enhanced, leading to a high multiplication and gain. We suggest that this simple approach can be adopted to implement very high sensitivity SiC APDs.
-676- Journal of the Korean Physical Society, Vol. 56, No. 2, February 2010 ACKNOWLEDGMENTS This work was supported by the Korean Research Foundation Grant funded by the Korean Government (MOEHRD, Basic Research Promotion Fund) (KRF- 2008-331-D00297). The author thanks S. Soloviev, A. Vertiatchikh, and P. M. Sandvik at GE Global Research for useful discussion. REFERENCES [1] Y. A. Goldberg, Semicond. Sci. Technol. 14, R41 (1999). [2] M. Razeghi and A. Rogalski, J. Appl. Phys. 79, 7433 (1996). [3] E. Monroy, F. Omnes and F. Calle, Semicond. Sci. Technol. 18, R33 (2003). [4] K.-S. Park, T. Kimoto and H. Matsunami, J. Korean Phys. Soc. 30, 123 (1997). [5] K.-S. Park, K.-S. Nahm, T. Kimoto and H. Matsunami, J. Korean Phys. Soc. 33, 86 (1998). [6] H.-Y. Cha and P. M. Sandvik, Japan. J. Appl. Phys. Part 1 47, 5423 (2008). [7] J. Cornu, Electron. Lett. 8, 169 (1972). [8] J. Cornu, IEEE Trans. Electron Dev. 20, 347 (1973). [9] available at http://www.ioffe.ru/sva/nsm/- Semicond/SiC/recombination.html. [10] K. Vassilevski, Int. J. High Speed Electron. Syst. 15, 899 (2005). [11] H.-Y. Cha, S. Soloviev, S. Zelakiewicz, P. Waldrab and P. M. Sandvik, IEEE Sensors J. 8, 233 (2008). [12] G. A. Shaw, A. M. Siegel and J. Model, Proc. of SPIE, 6231, 62310C.1 (2006). [13] A. A. Allerman, M. H. Crawfor, A. J. Fischer, K. H. A. Bogart, S. R. Lee, D. M. Follstaedt, P. P. Provencio and D. D. Koleske, J. Cryst. Growth 272, 227 (2004).